I recently prepared and recorded a video lecture on time series analysis and forecasting for University College of Professional Education in Wrocław.
The lecture covers key concepts and practical techniques used in working with temporal data, including preprocessing, resampling, smoothing, autocorrelation, stationarity, classical forecasting methods such as ARIMA and SARIMA, and modern transformer-based approaches for time series.
This project combined technical content creation, practical Python examples, and video-based teaching in a format designed for online education.
As part of the project, I also prepared a GitHub repository with Python examples: https://github.com/mzarnecki/time-series-analysis
1. time_series_analyze_with_pandas.ipynb
An introductory notebook on time series analysis using the pandas library. It demonstrates working with time indexes, data filtering, resampling, aggregation, and basic techniques for time-based data exploration and visualization.
2. time_series_arima_transformer.ipynb
This notebook compares classical and modern approaches to time series forecasting. It includes statistical models such as ARIMA as well as methods based on the Transformer architecture, making it possible to contrast traditional techniques with deep learning solutions.
3. prognose_google_stock_price.ipynb
An example of forecasting Google stock prices based on a historical time series. The notebook presents a typical financial data analysis pipeline: data preparation, trend visualization, train-test splitting, and building a model to predict future values.
4. siren_sound_analyse_and_classify.ipynb
This notebook presents the analysis of a siren audio signal and its classification. It includes elements of signal processing, feature extraction from audio data, and the use of machine learning models to recognize classes based on a time-dependent signal.
5. parkinson_fog_prediction.ipynb
This notebook shows an example of predicting Freezing of Gait (FoG) episodes in people with Parkinson’s disease based on time series data. It illustrates sequential data preparation, signal analysis, and the development of a model supporting a classification task in a medical data context.